At present, the cloud computing technology develops rapidly. As the machine scale of a data center of a public cloud is enlarged continuously, how to rapidly acquire and process monitoring data of machines in a whole network becomes a significant issue in cloud monitoring.
Conventionally, monitoring data is mainly acquired with two approaches: an agent-based approach and an agentless approach. With the agentless approach, a simple network management protocol (SNMP) protocol is required and the machines should support a relative interface. Hence, the agentless approach is not flexible and is rarely used. Conventionally, it is a mainstream to arrange an agent in the whole network and acquire the monitoring data with the agent-based approach. In a conventional agent-based approach, the agent periodically reports the monitoring data to a receiving server, while the receiving server can only receive passively. Therefore, the agent-based approach has disadvantages such as low data flexibility, low expandability, low manageability and poor performance.